48.7 million views from an AI character with no visible eyes: how to design an AI character people remember

The AI Character Derek Mercer: How I Built My Own After He Got 48.7 Million Views

The AI character Derek Mercer is a fictional "bodybuilder" on TikTok who never shows his eyes, and his first 12 videos combined got around 48.7 million views. This post summarizes and rewrites in Vietnamese the analysis of a creator who dissected that phenomenon and then built his own AI character. According to GHOST (@ghosstty_, 16/10/2026), there are four main points: why that character is so memorable, the character design process from reference sheets to video, costs around $155 for 10 videos, and why this wave might cool down fast. First is Derek's story, then the process, finally the costs and risks.

This post is based on "$48K and 48.7M Views From an AI Bodybuilder Whose Eyes You Never See. So I Built My Own AI Character" by GHOST (@ghosstty_) on X (16/10/2026), rewritten in Vietnamese by OPCAIOS, condensed and with notes added for content creators in Vietnam. The numbers about the accounts mentioned are self-counted or cited by the original author; OPCAIOS has not independently verified them.

Who is Derek Mercer and where did 48.7 million views come from?

Derek Mercer is a bodybuilder who never set foot in a gym: every video is AI-generated, and some captions even state straight that it was made with Higgsfield, an AI image and video creation platform. According to GHOST (2026), the TikTok account has around 170 thousand followers and 4.2 million likes, follows no one, and the person behind it has never shown their face, name, or voice.

Before trusting the numbers floating around, the author counted the views of the first 12 videos on the page themselves: total around 48.7 million views, with the highest video around 20.4 million. The money figure is separate. Some posts on X say the creator of Derek released a token on the pump.fun platform and collected around $48,000 in transaction fees; another account said $60,000. The author does not consider that income: token fees are more like a casino receipt than a salary, and disappear when the hype dies down. Views sit publicly on the page; anyone can check in 2 minutes.

Lesson 1: Face is not required, silhouette is

Silhouette recognition is the set of body features and clothing that help viewers remember a character even without seeing their face. According to GHOST (2026), most trend analysis posts about AI characters say face is everything, but Derek almost always has his emo bangs covering his eyes and is still the most recognizable character on the author's feed. For example, everything that makes Derek is under the hair:

• Black emo-style hair, long.
• Muscles so large they're unrealistic, turned into a joke.
• The same outfit in every video: gray shirt with blue graphic print, studded belt, wallet chain, black cargo pants.

Viewers recognize him from the shoulders down, from behind, at the size of a postage stamp. The captions even add a paradox: "Am I scary?", "I feel alone..." from a sad teenager in the body of someone who can flip a car. Viewers stick around to see which side wins. Face may not be needed. Silhouette is required.

Lesson 2: describe the character in one sentence

The one-sentence rule is a test that the character is done if the creator can describe it to friends in one sentence. According to GHOST (2026), the author set this rule before writing the first prompt, and Derek passed the test in five words: emo hair, unrealistic muscles, gray shirt. That sentence is the product; everything else is decoration.

For his own character, the author chose four features big enough to survive every camera angle: a hairstyle readable from across the street, eyes slightly too surprised, sparse front teeth showing every smile, and a suit in a color that acts like a logo. Add one more test from Derek's page: on a phone, each video box in the grid is only about thumb-width wide, and Derek is still recognizable across all 12 boxes. So every character design must be recognizable even when shrunk to a grid box.

Process: Three Reference Sheets, One Character Preserved

The author builds a character using a character reference sheet: a single image on a smooth gray background, showing full-body front view, side view, back view, plus two close-up face shots — so every video afterward uses the same "original version."

1. First sheet — discard immediately. The character is strange, but intentionally so: a hairstyle like a hat placed on the head for a photo, a blue velvet tuxedo that looks like a wedding party staff member, a face slanting toward caricature. This kind of "strange" people scroll past because it demands attention rather than earns it. The author keeps three things: round surprised eyes, sparse teeth, and a square hairstyle.
2. Second sheet — right face, wrong outfit. Rebuilt from scratch around the chosen features: black hair in a pillar shape standing nearly as tall as the face, eyes showing white around the pupils, sparse teeth with thin mustache and goatee. Face complete. But the cream vest, black shirt, yellow tie outfit is "very beautiful" — and very beautiful is something no one remembers. Shrink it to a single grid square, let the cream color blend into the gray background, leave only the hair visible.
3. Third sheet — edit, don't rewrite. This is the step that changes everything.

Why edit the image instead of writing a new prompt?

Editing (edit) means adjusting an existing reference sheet to keep the character intact and only swap the part you specify. A new prompt gives you a new person: even with the old description pasted in correctly, the model still "rolls the dice" again and hands you the character's cousin.

The author attaches the second sheet and requests keeping the face, eyes, teeth, goatee, skin, build, hair, background, lighting, composition unchanged — only change the outfit, identical from every angle. The new outfit is a turquoise silk vest woven with yellow flower and bee patterns, pink silk shirt with a large bow, gold chain and ring. The character is named Cornelius because turquoise, pink, and gold is a combination no one wears by accident. Shrink it to a single grid square, and Cornelius becomes a block of turquoise with the black hair pillar on top, recognizable before you see any facial detail, even from behind. The lesson packed into one line: design the face for close-ups, design the outfit for the thumbnail.

Give a character a "shop" before you give them a joke

Derek lifts weights — that's his entire world: a gym, a balcony, a body built for exactly one joke. The author wants Cornelius to own something, because a business gives you hundreds of pieces of free content. Cornelius becomes owner of a luxury fur shop — not a sales clerk. That one word changes how he moves: the clerk tries to convince you, the owner just shows you what's his and waits for you to understand.

The first video script:

• Camera angle: phone placed on a stone counter, angled slightly up — exactly how a real shop owner films inventory for social media. From this angle, the hair pillar is tall under the ceiling fan and he looks twice the size of the room.
• Action: introduce each item straight into the lens with the pride of someone who hand-selected the merchandise.
• Setting: coat racks in cream and honey color; two employees in tuxedos standing still in the background. No one reacts to the turquoise vest because he's the boss — the world treats him as completely normal, and that's precisely what makes him strange.
• Final beat: he pushes a large white fur piece straight into the lens until it fills the frame, leaving only his wide-open eyes peeking over the top — the only shot where the face almost disappears, and it's the shot people screenshot.

There's one more subtle reason: someone introducing a product to a camera is already an advertising format. If a brand collaboration comes later, the product placement already exists in the show — he holds the product up just the way he holds the fur.

Where to place the reference image: "reference", not "first frame"

The video was made using Seedance inside Higgsfield, with a full-body front-facing photo attached as a reference image, not as the first frame. The difference sounds technical but determines everything: using it as the opening frame, the video starts with a studio shot on a gray wall and dies of hallucination in half a second; using it as a reference, the model takes the face and outfit and places them in a completely new room.

Three sentences in the video prompt carry most of the results:

• "The phone never moves" — so the model doesn't invent a cameraman on its own.
• "The employee in the background stands completely still" — so the background doesn't steal attention.
• The push of a strand of hair into the lens at the end — it gives the video an ending. Most AI videos don't end, they just stop.

With dance videos, the author believes many videos in this wave take a viral dance performed by a real person, then use motion transfer tools (for example Kling Motion Control) to make the character move identically. The body language is borrowed from a real person, so it looks like a real person.

How much does it cost to make 10 AI character videos?

Making 10 complete AI character videos costs around $155 according to the most detailed public cost breakdown GHOST (2026) found: roughly $102 for the video part (320 seconds, about $0.32 per second), roughly $3.35 for 50 images, the rest for other AI calls, editing and music. Accounting for revisions, budget $200–300 for the first 10 videos.

The real cost is time. The same source estimates 30–50 hours for the first 10 videos when learning, and after one evening of building a reference sheet, the author found that number to be real. For example, if a content creator in Vietnam spends 2 hours a day, the first 10 videos take about 15 to 25 days, a calculation by OPCAIOS from the estimates above and not data from the author.

Is this AI character wave cooling down?

The AI character wave may be cooling down, according to the part many articles about Derek skip over. According to GHOST (2026), Derek's early videos got around 2 million views each, while new videos struggle to break 100 thousand. The mustachioed French character that opened this entire wave is said to have lost about 50% engagement after viewers discovered it was AI.

Novelty is the spark, not the fuel. Strange looks buy the first million views and almost nothing after that. What survives the spark is a viewing habit people want to encounter again: the same person, the same outfit, a small new situation each time. Cornelius' store next week sells hats, then perfume, then a very expensive pair of gloves. The author also raises two unresolved risks: dense patterns like tiny golden bees make the video model flicker when moving, and the formula is easy to copy within 1 week if Cornelius succeeds. Build a program, not a moment.

Clearly state "AI character" right from the start

A decision the author settled in advance: list "AI character" in the account introduction from the first post. Viewers click to find out how the character is made, and hiding it only makes comments angrier when they discover it.

For content creators in Vietnam, this is also a matter of compliance: TikTok, YouTube and Meta all require labeling content that looks like AI-generated real people, and when the character starts receiving ads, the ad content still must follow advertising rules like any real KOL — not being a fictional character doesn't mean it gets to make exaggerated product claims.

If You Want to Try Making an AI Character, How Do You Start?

The starting path is to make 10 videos by hand around a single character. According to GHOST (2026), the author proposes five steps for himself, from design to budget:

1. One character, describable in a single sentence, recognizable at the size of a grid square.
2. One reference sheet locked in, every change after that is image editing, not writing new prompts.
3. One repeatable setting (store, counter, kitchen...) instead of a one-off joke.
4. Ten videos made by hand, with a budget you can afford to lose without regret, around $200–300 by the estimate above.
5. No second character yet, no automation yet, no token until viewers demand the next episode of the first character.

As the author writes, an AI agent can post 100 videos a day, but it can't fix a character that no one wants to meet twice. After ten videos, you'll know whether a set of clothes can become a character or not.

Read more: Building a "faceless" TikTok/Reels channel for OPC; Step-by-step creating a TikTok/Shorts channel automated with HeyGen, Suno and CapCut from A-Z.

Source: GHOST (@ghosstty_) on X, 16/10/2026 — rewritten and condensed by OPCAIOS; account figures and costs per original author.